dsh-vibe-math 0.3.4 → 0.3.6

package/README.md

@@ -15,6 +15,24 @@
 
 ---
 
+## 🧩 架构图（v1 + v2）
+
+> 静态架构图；完整流程说明见 [docs/架构图.md](docs/架构图.md)；可编辑生成脚本：[v1](docs/generate_framework_diagram.py) / [v2](docs/generate_framework_diagram_v2.py)。
+
+### Vibe Math V1（经典流水线）
+
+![Vibe Math V1 架构图](示例图/框架图-v1.png)
+
+**一句话流水线**：`qs.csv` → Brainstorm 拆方向 → 每方向一个 Solver 多轮迭代（卡死则 Derive 派生新方向）→ 输出拆成最小验证单元 → ≥3 个 Verifier 独立审查 → 辩论 → 裁决 → 通过晋升 `Verified/` → Decider 回写 `qs.csv`；全程状态落盘 `VibeMath_State/`，`resume` 断点续跑，`manual` 模式在派发/裁决/晋升处挂起人工决策。
+
+### Vibe Math V2（新架构 · 概率驱动）
+
+![Vibe Math V2 架构图](示例图/框架图-v2.png)
+
+**一句话流水线**：`qs.json` 按优先级取问题 → Explorer 拆方向（全死路则重派生）→ 每方向一个 Solver 多轮迭代（引理进 `Propos/`、解法回 `qs.json`，概率均 <1）→ 调度器选 r（命题 / 命题+证明·证伪 / 问题+解法）派 ≥3 验证器独立审查→辩论→裁决 → 概率=1 自动收口（问题 solved、命题 1/0，优先级置 `never`）；全程状态落盘，`resume` 断点续跑，`reportMode` 可 file/push/both 汇报。
+
+---
+
 ## ✨ 功能特色
 
 - **多代理自动求解**：主代理把问题交给调度器，调度器派发 brainstorm / solver / verifier / decider（v1）或 explorer / solver / verifier（v2）等子代理协同求解，**你无需逐节点手操**。
@@ -43,6 +61,7 @@ dsh plugin --profile <你的 profile> add github:ChongCyrus/Vibe-Mathematics
 
 安装时插件会自动把两个 preset 写入 `~/.dsh/.agent-presets/`：`vibe-math-v1/` 与 `vibe-math-v2/`。
 之后新建会话，预设选择器里选择 **Vibe Math**（v1）或 **Vibe Math V2**（v2）即可。
+**升级包版本后重启 DSH，未手动改过的 preset 文件会自动更新到新版本**（细节见文末「v2」安装器说明）。
 
 ### 方式 B：作为 agent preset 手动安装
 
@@ -92,116 +111,6 @@ dsh plugin --profile <你的 profile> add github:ChongCyrus/Vibe-Mathematics
 
 ---
 
-## 🧩 架构图 · v1（经典）
-
-> 静态 PNG 预览 + 可编辑 Mermaid 图源（GitHub 原生渲染）；完整流程说明见 [docs/架构图.md](docs/架构图.md)。
-
-![Vibe Math V1 架构图](示例图/框架图-v1.png)
-
-```mermaid
-flowchart TB
-    subgraph L1["👤 交互层"]
-        U["😀 用户（自然语言）"]
-        M["🤖 主代理（助手 + 汇报者）<br/>翻译需求 · 汇报进度 · 问答配置<br/>不求解 · 不调度"]
-    end
-
-    subgraph L2["⚙️ 调度层（插件代码）"]
-        SCHED["调度器 Scheduler<br/>唯一文件写者<br/>读 qs.csv · 派发子代理 · 推进状态机"]
-    end
-
-    subgraph L3["🧠 子代理层（continuable 持久会话）"]
-        BRAIN["🧭 Brainstorm<br/>拆解多个求解方向"]
-        SOLV["✍️ Solver × N<br/>逐方向多轮迭代"]
-        DERI["🔀 Derive<br/>死路时派生新方向"]
-        VERI["🔬 Verifier × ≥3<br/>独立审查 → 辩论 → 裁决"]
-        DECI["⚖️ Decider<br/>判定是否解决"]
-    end
-
-    subgraph L4["💾 数据层 · VibeMath/Projects/&lt;项目&gt;/"]
-        QS["📋 qs.csv"]
-        PEND["📥 Pending_Verification/"]
-        UNDR["📂 Under_Verification/"]
-        TVAL["✅ Temp_Validated/"]
-        KNOW["📚 Verified/ 可信知识库"]
-        STATE["🗄️ Progress_Logs/ · VibeMath_State/"]
-    end
-
-    U -->|"求解 XX / 查进度 / 干预"| M
-    M -->|"vibe_math_* 工具"| SCHED
-
-    SCHED -->|"① 派发"| BRAIN
-    BRAIN -->|"多个大相径庭的方向"| SCHED
-    SCHED -->|"② 每方向一个"| SOLV
-    SOLV -->|"结构化结果 JSON"| SCHED
-    SOLV -.->|"迭代至卡死"| DERI
-    DERI -->|"派生 1~3 个新方向"| SCHED
-    SCHED -->|"③ 写盘"| PEND
-    SCHED -->|"④ 拆解为最小验证单元"| UNDR
-    SCHED -->|"⑤ 派发 ≥3 个"| VERI
-    VERI -->|"裁决 JSON"| SCHED
-    SCHED -->|"⑥ 通过"| TVAL
-    SCHED -->|"⑦ 晋升"| KNOW
-    SCHED -->|"⑧ 派发"| DECI
-    DECI -->|"⑨ 已解决 → 回写"| QS
-    SCHED <-->|"读取问题"| QS
-    SCHED <-->|"状态落盘 / 断点恢复"| STATE
-```
-
-**一句话流水线**：`qs.csv` → Brainstorm 拆方向 → 每方向一个 Solver 多轮迭代（卡死则 Derive 派生新方向）→ 输出拆成最小验证单元 → ≥3 个 Verifier 独立审查 → 辩论 → 裁决 → 通过晋升 `Verified/` → Decider 回写 `qs.csv`；全程状态落盘 `VibeMath_State/`，`resume` 断点续跑，`manual` 模式在派发/裁决/晋升处挂起人工决策。
-
----
-
-## 🧩 架构图 · v2（新架构 · 概率驱动）
-
-> 静态 PNG 预览 + 可编辑 Mermaid 图源（GitHub 原生渲染）；生成脚本见 [docs/generate_framework_diagram_v2.py](docs/generate_framework_diagram_v2.py)。
-
-![Vibe Math V2 架构图](示例图/框架图-v2.png)
-
-```mermaid
-flowchart TB
-    subgraph L1["👤 交互层"]
-        U["😀 用户（自然语言）"]
-        M["🤖 主代理（助手 + 汇报者）<br/>翻译需求 · 汇报进度 · 问答配置<br/>不求解 · 不调度"]
-    end
-
-    subgraph L2["⚙️ 调度层（插件代码 · 概率驱动）"]
-        SCHED["调度器 Scheduler<br/>唯一文件写者 · 按优先级调度<br/>explorer → 逐方向 solver → 验证器 → 收口<br/>断点持久化 · 人工决策门"]
-    end
-
-    subgraph L3["🧠 子代理层（continuable 持久会话）"]
-        EXP["🧭 Explorer<br/>拆方向 / 全死路时重派生（去死路 · 并集）"]
-        SOLV["✍️ Solver × N<br/>agent_self_iteration<br/>逐方向多轮迭代"]
-        VERI["🔬 Verifier × ≥3<br/>独立审查 → 辩论 → 裁决"]
-    end
-
-    subgraph L4["💾 数据层 · VibeMath/Projects/&lt;项目&gt;/"]
-        QS["📋 qs/qs.json<br/>问题 + 解法（正确概率）"]
-        PRP["📚 Propos/&lt;分类&gt;_Propos.json<br/>命题（布尔估计 · 证明/证伪列表）"]
-        VERD["✅ Verified/<br/>定论事实（0/1）"]
-        LOGS["🗄️ Verification_logs/ · Progress_Logs/<br/>· Reliable/ · VibeMath_State/"]
-    end
-
-    U -->|"求解 / 查进度 / 干预"| M
-    M -->|"vibe_math_* 工具"| SCHED
-
-    SCHED -->|"① 按优先级取问题"| QS
-    SCHED -->|"② 派 Explorer"| EXP
-    EXP -->|"方向集 M_q"| SCHED
-    SCHED -->|"③ 每方向一个"| SOLV
-    SOLV -->|"引理 → Propos（概率&lt;1）"| PRP
-    SOLV -->|"解法 → qs（概率&lt;1）"| QS
-    SOLV -.->|"卡死 → 重派生"| EXP
-    SCHED -->|"④ 选 r 派 ≥3 验证器"| VERI
-    VERI -->|"⑤ 裁决 0/1/中间"| SCHED
-    SCHED -->|"⑥ 概率=1 → 收口/晋升"| PRP
-    SCHED -->|"已解决 / 优先级 never"| QS
-    SCHED -->|"晋升定论"| VERD
-    SCHED <-->|"辩论记录 / 状态落盘"| LOGS
-```
-
-**一句话流水线**：`qs.json` 按优先级取问题 → Explorer 拆方向（全死路则重派生）→ 每方向一个 Solver 多轮迭代（引理进 `Propos/`、解法回 `qs.json`，概率均 <1）→ 调度器选 r（命题 / 命题+证明·证伪 / 问题+解法）派 ≥3 验证器独立审查→辩论→裁决 → 概率=1 自动收口（问题 solved、命题 1/0，优先级置 `never`）；全程状态落盘，`resume` 断点续跑，`reportMode` 可 file/push/both 汇报。
-
----
 
 ## 📁 目录结构
 
@@ -278,9 +187,9 @@ flowchart TB
 | `vibe_math_setup` | 返回参数 schema（交互式配置用） |
 | `vibe_math_save_settings` | 把当前参数存成新默认 |
 | `vibe_math_template` | 生成默认参数模板文件 |
-| `vibe_math_new_project` / `set_project` / `list_projects` | 项目管理 |
+| `vibe_math_new_project` / `vibe_math_set_project` / `vibe_math_list_projects` | 项目管理 |
 | `vibe_math_list_decisions` / `decide` | 查看 / 裁决人工决策 |
-| `vibe_math_list_agents` / `message_agent` / `interrupt_agent` | 查看 / 发消息 / 中断子代理 |
+| `vibe_math_list_agents` / `vibe_math_message_agent` / `vibe_math_interrupt_agent` | 查看 / 发消息 / 中断子代理 |
 
 斜杠命令（与工具等价）：`/vibe start|resume|pause|abort|status|report|mode <auto|manual>|setup|save|template [global|project]|add <id> <desc>|add-proposition <id> <概述>|list-propositions|project [list|new <name>|<name>]|decisions|agents`
 
@@ -414,6 +323,7 @@ flowchart TB
 | `mode` | `auto` | `auto` / `manual` |
 | `maxParallelThreshold` | 4 | 全局最大并发子代理轮数（新派发前须 active < 阈值） |
 | `solverMaxRounds` | 3 | 每个求解方向最大迭代轮数（agent_self_iteration 上限） |
+| `directionsPerSolver` | 1 | 每个 solver 提示词附带的方向数量（1 = 只看自己方向、互不干扰；>1 = 附带其他活跃方向摘要用于协调） |
 | `verifierCount` | 3 | 每个验证对象的独立验证器数量 |
 | `debateMaxRounds` | 5 | 验证辩论（交流群）最大轮数 |
 | `verdictMode` | `flat` | `flat` = 均衡机制（不一致判 0.5）/ `forced` = 强制裁决（历史准确率+严谨性加权） |
@@ -456,7 +366,7 @@ flowchart TB
 - manual 模式在第一个未决关键节点暂停整条主循环。
 
 **v2**：
-- 安装器只**补装缺失**的 preset 文件，不覆盖你已编辑的预设（删除对应目录即可重新安装）。
+- 安装器带**版本化自动更新**：每次 DSH 启动时对比包版本与 `<presetRoot>/.vibe-math-installed.json` 记录——版本升级会自动替换**未被手动修改**的 preset 文件（哈希一致才覆盖）；你改过的文件会被保留并在日志中提示。无记录的老安装首次会一次性刷新到当前版本。想强制全量重装：删除 `~/.dsh/.agent-presets/vibe-math-v1` 与 `vibe-math-v2` 目录后重启 DSH。
 - `flat` 裁决在辩论不一致时直接判 `0.5`；`forced` 按历史准确率+置信度加权。
 - `never` 优先级的问题/命题**永不调度**，且不阻塞严格终止（视为主动弃权）。
 - 两个 preset 文件互相独立、可共存；同一会话同时只能选一个预设。

package/installer.js

@@ -1,51 +1,149 @@
-// dsh-vibe-math merged bundle installer.
+// dsh-vibe-math merged bundle installer — VERSIONED AUTO-UPDATE.
 // When this bundle is installed (e.g. `dsh plugin add dsh-vibe-math` or from the
 // dsh-market), this plugin copies BOTH agent presets out of the package into the
 // DSH preset root, so the user immediately gets two presets in the picker:
 //   vibe-math-v1/  (classic pipeline architecture)
 //   vibe-math-v2/  (new probability-driven architecture)
-// Copy is per-file and only installs MISSING files (idempotent; never clobbers
-// a preset the user already edited — delete the dir to force a reinstall).
-import { existsSync, mkdirSync, readFileSync, writeFileSync } from 'node:fs'
+//
+// UPDATE POLICY (state recorded in <presetRoot>/.vibe-math-installed.json):
+//   - baseline (no state file — e.g. upgrading from an installer that predates
+//     this mechanism): every existing owned file is refreshed to the current
+//     package version and recorded as package-owned (user policy: auto-update
+//     old installs; any manual edits made before this baseline are overwritten
+//     once — from then on edits are protected).
+//   - upgrade (recorded version != current package.json version): every owned
+//     file that is byte-identical to the previously installed copy (i.e. NOT
+//     user-edited since) is overwritten with the new version; user-edited files
+//     are preserved and reported via the logger.
+//   - same version: no-op (idempotent). Missing files are ALWAYS restored.
+//   - force a full refresh at any time: delete the preset dirs and restart DSH.
+import { existsSync, mkdirSync, readFileSync, renameSync, writeFileSync } from 'node:fs'
+import { createHash } from 'node:crypto'
 import { homedir } from 'node:os'
 import { dirname, join } from 'node:path'
 import { fileURLToPath } from 'node:url'
 
 export const name = 'vibe-math-preset-installer'
 
+const PRESETS = [
+  {
+    src: 'vibe-math-v1',
+    dst: 'vibe-math-v1',
+    files: ['agent.cordis.yml', 'preset.yml', 'vibe-math.js', '实现方案-多代理数学问题求解与验证框架.md'],
+  },
+  {
+    src: 'vibe-math-v2',
+    dst: 'vibe-math-v2',
+    files: ['agent.cordis.yml', 'preset.yml', 'vibe-math-v2.js', '实现方案.md'],
+  },
+]
+
+const STATE_FILE = '.vibe-math-installed.json'
+
+function sha256(buf) { return createHash('sha256').update(buf).digest('hex') }
+
+function readState(path) {
+  try {
+    const raw = readFileSync(path, 'utf8')
+    const obj = JSON.parse(raw)
+    if (obj && typeof obj === 'object' && obj.files && typeof obj.files === 'object') return obj
+  } catch (e) { /* missing or corrupt — treat as no state (baseline) */ }
+  return null
+}
+
+function writeState(path, state) {
+  try {
+    const tmp = path + '.tmp'
+    writeFileSync(tmp, JSON.stringify(state, null, 2) + '\n', 'utf8')
+    renameSync(tmp, path)
+  } catch (e) {
+    // best-effort: state persistence failure must not break the copy step
+  }
+}
+
 export function apply(ctx) {
   const logger = ctx && ctx.logger
   try {
     const dshHome = process.env.DSH_HOME || join(homedir(), '.dsh')
     const here = dirname(fileURLToPath(import.meta.url))
-    const targets = [
-      {
-        src: join(here, 'vibe-math-v1'),
-        dst: join(dshHome, '.agent-presets', 'vibe-math-v1'),
-        files: ['agent.cordis.yml', 'preset.yml', 'vibe-math.js', '实现方案-多代理数学问题求解与验证框架.md'],
-      },
-      {
-        src: join(here, 'vibe-math-v2'),
-        dst: join(dshHome, '.agent-presets', 'vibe-math-v2'),
-        files: ['agent.cordis.yml', 'preset.yml', 'vibe-math-v2.js', '实现方案.md'],
-      },
-    ]
-    let installed = 0
-    for (const t of targets) {
-      if (!existsSync(t.src)) continue
-      mkdirSync(t.dst, { recursive: true })
-      for (const f of t.files) {
-        const s = join(t.src, f)
-        const d = join(t.dst, f)
-        if (!existsSync(s) || existsSync(d)) continue
-        writeFileSync(d, readFileSync(s))
-        installed += 1
+    const presetRoot = join(dshHome, '.agent-presets')
+    const stateFile = join(presetRoot, STATE_FILE)
+
+    // current package version (the source of truth for "is this an upgrade?")
+    let pkgVersion = ''
+    try { pkgVersion = String((JSON.parse(readFileSync(join(here, 'package.json'), 'utf8')).version) || '') } catch (e) { pkgVersion = '' }
+
+    const state = readState(stateFile)
+    const prevFiles = (state && state.files) || {}
+    const isUpgrade = state !== null && pkgVersion !== '' && state.version !== pkgVersion
+    const isBaseline = state === null // no recorded history → refresh everything (user policy: auto-update old installs)
+
+    const nextFiles = {}
+    let installed = 0, updated = 0, kept = 0
+    const keptList = []
+
+    for (const p of PRESETS) {
+      const srcDir = join(here, p.src)
+      const dstDir = join(presetRoot, p.dst)
+      if (!existsSync(srcDir)) continue
+      mkdirSync(dstDir, { recursive: true })
+      for (const f of p.files) {
+        const s = join(srcDir, f)
+        const d = join(dstDir, f)
+        if (!existsSync(s)) continue
+        const key = p.src + '/' + f
+        const cur = readFileSync(s)
+        const curHash = sha256(cur)
+        if (!existsSync(d)) {
+          // missing file: always restore, whatever the version
+          writeFileSync(d, cur)
+          installed += 1
+          nextFiles[key] = { hash: curHash, provenance: 'package' }
+          continue
+        }
+        const destHash = sha256(readFileSync(d))
+        if (isBaseline) {
+          // no recorded history: refresh to the current package (one-time; edits
+          // made before this mechanism are overwritten, later edits are protected)
+          if (destHash === curHash) { nextFiles[key] = { hash: curHash, provenance: 'package' } }
+          else { writeFileSync(d, cur); updated += 1; nextFiles[key] = { hash: curHash, provenance: 'package' } }
+          continue
+        }
+        const prev = prevFiles[key]
+        const prevRec = (prev && typeof prev === 'object') ? prev : { hash: prev, provenance: 'package' }
+        const prevProv = (prevRec.provenance === 'user') ? 'user' : 'package' // 未知来源按包文件处理
+        if (prevProv === 'package' && destHash === prevRec.hash) {
+          // 包文件且未被改动 → 可安全升级（内容相同则跳过写入）
+          if (destHash !== curHash) { writeFileSync(d, cur); updated += 1 }
+          nextFiles[key] = { hash: curHash, provenance: 'package' }
+        } else if (prevProv === 'user') {
+          // 用户持有 → 永不覆盖
+          kept += 1
+          if (isUpgrade) keptList.push(key + ' (用户持有)')
+          nextFiles[key] = { hash: destHash, provenance: 'user' }
+        } else {
+          // 包文件但自上次安装后已被用户改动
+          kept += 1
+          if (isUpgrade) keptList.push(key + ' (已修改)')
+          nextFiles[key] = { hash: destHash, provenance: 'user' }
+        }
       }
     }
-    if (installed > 0) {
-      logger?.info?.('[dsh-vibe-math] installed ' + installed + ' preset file(s) — presets vibe-math-v1 & vibe-math-v2 are now available in the preset picker (new session)')
+
+    writeState(stateFile, { version: pkgVersion, files: nextFiles, updatedAt: Date.now() })
+
+    if (isUpgrade) {
+      logger?.info?.('[dsh-vibe-math] preset auto-update: version ' + (state.version || '(none)') + ' → ' + pkgVersion +
+        ' — 新增 ' + installed + ' 个文件，更新 ' + updated + ' 个文件' +
+        (kept > 0 ? '，保留 ' + kept + ' 个未覆盖文件（' + keptList.join('; ') + '）' : '') +
+        '。新版本 preset 将在新会话生效。')
+    } else if (isBaseline) {
+      logger?.info?.('[dsh-vibe-math] preset baseline: refreshed ' + (installed + updated) + ' file(s) to v' + pkgVersion +
+        ' — 已启用自动更新（后续版本升级将自动替换未被手动修改的 preset 文件）。')
+    } else if (installed > 0) {
+      logger?.info?.('[dsh-vibe-math] restored ' + installed + ' missing preset file(s)')
     }
   } catch (err) {
-    logger?.warn?.('[dsh-vibe-math] preset install failed: %s', String((err && err.message) || err))
+    logger?.warn?.('[dsh-vibe-math] preset install/update failed: %s', String((err && err.message) || err))
   }
 }

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "dsh-vibe-math",
   "description": "Multi-agent mathematical problem-solving & verification frameworks for DeepSeek Harness — TWO agent presets in one install: vibe-math-v1 (classic pipeline: brainstorm → solver iteration → multi-verifier debate → Verified) and vibe-math-v2 (new probability-driven architecture: qs.json + Propos knowledge base + explorer→solver→review/debate verdict). Installing this bundle auto-installs both presets into the DSH preset root.",
-  "version": "0.3.4",
+  "version": "0.3.6",
   "type": "module",
   "main": "installer.js",
   "exports": {

package/vibe-math-v2/vibe-math-v2.js

@@ -34,6 +34,7 @@ export function apply(ctx) {
     mode: 'auto',                 // auto | manual
     maxParallelThreshold: 4,      // concurrency gate: active turns < this
     solverMaxRounds: 3,           // per-direction iteration cap (spec example)
+    directionsPerSolver: 1,       // 每个 solver 提示词附带的方向数量（1 = 只看自己方向）
     verifierCount: 3,             // independent reviewers per verification
     debateMaxRounds: 5,           // debate round cap (spec example)
     verdictMode: 'flat',          // flat = 均衡机制(0.5) | forced = 强制裁决(weighted)
@@ -119,6 +120,7 @@ export function apply(ctx) {
     { name: 'mode', type: 'enum', options: ['auto', 'manual'], description: 'auto = 无人值守自动通过关键节点；manual = 关键节点挂起人工决策', suggestion: 'auto' },
     { name: 'maxParallelThreshold', type: 'integer', description: '全局最大并发子代理轮数（新派发前须满足 active < 阈值）', suggestion: 4 },
     { name: 'solverMaxRounds', type: 'integer', description: '每个求解方向的最大迭代轮数（agent_self_iteration 上限）', suggestion: 3 },
+    { name: 'directionsPerSolver', type: 'integer', description: '每个 solver 提示词附带的方向数量：1 = 只看自己方向（互不干扰）；>1 = 额外附带其他活跃方向摘要用于协调', suggestion: 1 },
     { name: 'verifierCount', type: 'integer', description: '每个验证对象的独立验证器数量', suggestion: 3 },
     { name: 'debateMaxRounds', type: 'integer', description: '验证辩论（交流群）最大轮数', suggestion: 5 },
     { name: 'verdictMode', type: 'enum', options: ['flat', 'forced'], description: 'flat = 均衡机制（不一致直接判 0.5）；forced = 强制裁决（按历史准确率+严谨性加权）', suggestion: 'flat' },
@@ -164,7 +166,7 @@ export function apply(ctx) {
   // ================= settings =================
   function sanitizeParams(obj) {
     const out = {}
-    const intFields = ['maxParallelThreshold', 'solverMaxRounds', 'verifierCount', 'debateMaxRounds', 'solverMaxToolCalls', 'verifierMaxToolCalls', 'reportIntervalMs', 'tickIntervalMs', 'activityLogCap', 'maxExplorerRetries']
+    const intFields = ['maxParallelThreshold', 'solverMaxRounds', 'directionsPerSolver', 'verifierCount', 'debateMaxRounds', 'solverMaxToolCalls', 'verifierMaxToolCalls', 'reportIntervalMs', 'tickIntervalMs', 'activityLogCap', 'maxExplorerRetries']
     const numFields = ['promoteValueThreshold']
     const arrayFields = ['solverToolAllow', 'solverToolDeny', 'verifierToolAllow', 'verifierToolDeny']
     for (const k of Object.keys(DEFAULT_PARAMS)) {
@@ -397,6 +399,16 @@ export function apply(ctx) {
       (d.lessons && d.lessons.length ? ' | lessons: ' + d.lessons.join('; ') : '') +
       (d.blockers && d.blockers.length ? ' | blockers: ' + d.blockers.join('; ') : '')
   }
+  // 每个 solver 默认只看自己方向（互不干扰）：round>1 时带上自己的历史进度，
+  // directionsPerSolver>1 时再附带其他活跃方向摘要用于协调（总数不超过该参数）。
+  function buildSolverContext(all, own, round, perSolver) {
+    const out = []
+    const n = Math.max(1, Number(perSolver) || 1)
+    if (round > 1) out.push(own)
+    const others = all.filter(function (d) { return d.id !== own.id && d.status === 'active' })
+    for (let i = 0; i < others.length && out.length < n; i++) out.push(others[i])
+    return out.map(directionSummary).join('\n')
+  }
   function solverPrompt(q, dir, round, progressText) {
     let head = solverPersonaText() + 'You are a dedicated solver agent working ONE solution direction of a math problem (agent_self_iteration).\n\n'
     head += 'PROBLEM (id: ' + q.id + '): ' + q.概述 + '\nDIRECTION: ' + dir.title + ' (method: ' + dir.method + '; core assumption: ' + dir.core_assumption + ')\nROUND: ' + round + ' of ' + params.solverMaxRounds + '\n'
@@ -688,7 +700,7 @@ export function apply(ctx) {
           if (dir.status === 'success' || dir.status === 'dead-end') continue
           const running = Object.keys(agentRegistry).some(function (cid) { const m = agentRegistry[cid]; return m && m.qid === q.id && m.direction === dir.id && m.role === 'solver' })
           if (running) continue
-          const progressText = prog.directions.map(directionSummary).join('\n')
+          const progressText = buildSolverContext(prog.directions, dir, 1, params.directionsPerSolver)
           const promptText = solverPrompt(q, dir, 1, progressText)
           const r = await maybeGate('spawn', 'solver for problem ' + q.id + ' direction ' + dir.id, { label: 'solver:' + q.id + ':' + dir.id, promptText: promptText, meta: { role: 'solver', qid: q.id, direction: dir.id, round: 1, description: q.概述 } }, async function (d) { await spawnChild(d.label, d.promptText, d.meta); return { spawned: true } })
           if (r && r.gated) return
@@ -743,7 +755,7 @@ export function apply(ctx) {
           delete agentRegistry[childId]
           logActivity('solver', qid + '/' + dirId + ' dead-end (success without solution)')
         } else {
-          const progressText = prog.directions.map(directionSummary).join('\n')
+          const progressText = buildSolverContext(prog.directions, dir, meta.round + 1, params.directionsPerSolver)
           await followupChild(childId, solverPrompt(q, dir, meta.round + 1, progressText))
           agentRegistry[childId].round = meta.round + 1
           dir.round = meta.round + 1
@@ -755,7 +767,7 @@ export function apply(ctx) {
       delete agentRegistry[childId]
       logActivity('solver', qid + '/' + dirId + ' dead-end: ' + dir.dead_end_reason)
     } else {
-      const progressText = prog.directions.map(directionSummary).join('\n')
+      const progressText = buildSolverContext(prog.directions, dir, meta.round + 1, params.directionsPerSolver)
       if (!scheduler.running) {
         // paused/aborted: stop the follow-up chain; keep the direction active for resume
         delete agentRegistry[childId]
@@ -1095,7 +1107,7 @@ export function apply(ctx) {
   registerTool('vibe_math_status', 'Show scheduler status, params, active agents, projects, and recent activity.', objParams({}), async function () { return await getStatus() })
   registerTool('vibe_math_report', 'Return the full progress report and write it to Progress_Logs/report.json.', objParams({}), async function () { await maybeWriteReport(true); return await buildReport() })
   registerTool('vibe_math_set_mode', 'Switch between manual and auto (preset) mode. Switching to auto auto-resolves any pending manual decisions.', objParams({ mode: { type: 'string', enum: ['manual', 'auto'] } }, ['mode']), async function (args) { params.mode = args.mode; await saveAll(); if (params.mode === 'auto') await autoResolvePending(); return { ok: true, mode: params.mode } })
-  registerTool('vibe_math_set_params', 'Update scheduler parameters (partial).', objParams({ maxParallelThreshold: { type: 'integer' }, solverMaxRounds: { type: 'integer' }, verifierCount: { type: 'integer' }, debateMaxRounds: { type: 'integer' }, verdictMode: { type: 'string', enum: ['flat', 'forced'] }, reportMode: { type: 'string', enum: ['file', 'push', 'both'] }, promoteValueThreshold: { type: 'number' }, priorityAdjust: { type: 'string', enum: ['none', 'deadend-deprioritize', 'survival-map'] }, proposPriorityAdjust: { type: 'string', enum: ['none', 'progress-graded'] }, provider: { type: 'string' }, model: { type: 'string' }, solverPersona: { type: 'string' }, verifierPersona: { type: 'string' }, solverToolAllow: { type: 'array', items: { type: 'string' } }, solverToolDeny: { type: 'array', items: { type: 'string' } }, verifierToolAllow: { type: 'array', items: { type: 'string' } }, verifierToolDeny: { type: 'array', items: { type: 'string' } }, solverMaxToolCalls: { type: 'integer' }, verifierMaxToolCalls: { type: 'integer' }, reportIntervalMs: { type: 'integer' }, tickIntervalMs: { type: 'integer' }, activityLogCap: { type: 'integer' }, maxExplorerRetries: { type: 'integer' } }), async function (args) { params = Object.assign({}, params, sanitizeParams(args)); await saveAll(); return { ok: true, params: params } })
+  registerTool('vibe_math_set_params', 'Update scheduler parameters (partial).', objParams({ maxParallelThreshold: { type: 'integer' }, solverMaxRounds: { type: 'integer' }, verifierCount: { type: 'integer' }, debateMaxRounds: { type: 'integer' }, verdictMode: { type: 'string', enum: ['flat', 'forced'] }, reportMode: { type: 'string', enum: ['file', 'push', 'both'] }, promoteValueThreshold: { type: 'number' }, priorityAdjust: { type: 'string', enum: ['none', 'deadend-deprioritize', 'survival-map'] }, proposPriorityAdjust: { type: 'string', enum: ['none', 'progress-graded'] }, provider: { type: 'string' }, model: { type: 'string' }, solverPersona: { type: 'string' }, verifierPersona: { type: 'string' }, solverToolAllow: { type: 'array', items: { type: 'string' } }, solverToolDeny: { type: 'array', items: { type: 'string' } }, verifierToolAllow: { type: 'array', items: { type: 'string' } }, verifierToolDeny: { type: 'array', items: { type: 'string' } }, solverMaxToolCalls: { type: 'integer' }, verifierMaxToolCalls: { type: 'integer' }, reportIntervalMs: { type: 'integer' }, tickIntervalMs: { type: 'integer' }, activityLogCap: { type: 'integer' }, maxExplorerRetries: { type: 'integer' }, directionsPerSolver: { type: 'integer' } }), async function (args) { params = Object.assign({}, params, sanitizeParams(args)); await saveAll(); return { ok: true, params: params } })
   registerTool('vibe_math_setup', 'Return the interactive parameter schema for guided configuration.', objParams({}), async function () { const list = PARAM_SCHEMA.map(function (p) { const out = Object.assign({}, p); out.current = params[p.name]; out.default = DEFAULT_PARAMS[p.name]; return out }); return { ok: true, parameters: list, saveTo: frameworkRoot() + '/vibe_math_setting.json' } })
   registerTool('vibe_math_save_settings', 'Write the current params to vibe_math_setting.json (JSON with comments) as new defaults.', objParams({}), async function () { return await saveSettings() })
   registerTool('vibe_math_template', 'Create a fresh vibe_math_setting.json template (with defaults + comments) in the workspace (global) or current project folder.', objParams({ where: { type: 'string', enum: ['global', 'project'] } }), async function (args) { return await createTemplate((args && args.where) || 'global') })